Skip to content
Model Pareto

Nanbeige4.1-3B

NanbeigeOpen weightsartificialanalysis.ai ↗
Input
—
Output
—
Speed
~110tok/sest.
Context
256Ktokens

Long-tail import from Artificial Analysis (2026-09-24); AA variant: Nanbeige4.1-3B. Price = AA median host price.

Rankings

Self-hostingEstimated

How we estimate →

What it would cost to run these open weights yourself on rented GPUs. No API sells this model, so this is its price on the chart.

Hardware
1× B200 180GB
BF16 weights
Throughput
~15,920 tok/s
many requests batched
Price
$0.047–0.42 /M tok
busy → light use
On your own machine
RTX 5090 32GB (int4, 300+ tok/s single-stream)
single consumer GPU or Mac
Assumptions (5)
  • BF16 weights (3.93B params) + 25% KV-cache headroom ≈ 10 GB
  • 1× B200 180GB at $5.98–$14.24/GPU-hour on-demand (2026-09-24)
  • ~15,920 output tok/s aggregate at batch 256 (bandwidth-bound)
  • Blended 3:1 input:output; prefill ~135,258 tok/s
  • Low = 75% utilization at the low GPU price; high = 20% at the high price

Benchmark scores

~ italic, dashed = no published score yet, estimated from related benchmarks.

Writing & chat

MMMLUEstimated
~last of 5 (estimated)below measured range (<81.3%)
  1. #1Gemini 3.1 Pro92.6%
  2. #2Gemma 4 31B88.4%
  3. #4gpt-oss-120b81.3%
  4. ~#5Nanbeige4.1-3BEstimated<81.3%
No published score — estimated from related benchmarks
~#46 of 57 (estimated)~1365 Elo
  1. #1Gemini 4 Argon1525
  2. #2Claude Opus 5.51507
  3. #45Trinity Large Thinking1369
  4. ~#46Nanbeige4.1-3BEstimated1365
  5. #46INTELLECT-31356
No published score — estimated from related benchmarks
~#22 of 26 (estimated)~1437 Elo
  1. #1GPT-6 Astra2173
  2. #2Claude Fable 5.12162
  3. #21Gemini 3.1 Pro1491
  4. ~#22Nanbeige4.1-3BEstimated1437
  5. #22Mistral Large 31412
No published score — estimated from related benchmarks

Hard reasoning

Agents

~#12 of 15 (estimated)~56.4%
  1. #1Kimi K384.8%
  2. #2Qwen3.8 27B84.3%
  3. #11Muse Glimmer65.9%
  4. ~#12Nanbeige4.1-3BEstimated56.4%
  5. #12Nemotron 3.5 Super VL NVFP452.3%
No published score — estimated from related benchmarks
OSWorld 2.0Estimated
~#3 of 5 (estimated)~53.8%
  1. #1GLM-5.3 Flash59.1%
  2. #2Kimi K358.3%
  3. ~#3Nanbeige4.1-3BEstimated53.8%
  4. #3GPT-5.6 Terra50.2%
No published score — estimated from related benchmarks
BrowseCompEstimated
~#21 of 22 (estimated)~40.1%
  1. #1Atria Dawn Preview92.5%
  2. #2GPT-6 Astra91.5%
  3. #20Nemotron 3 Ultra44.4%
  4. ~#21Nanbeige4.1-3BEstimated40.1%
  5. #21Mistral Small 421.3%
No published score — estimated from related benchmarks
~#38 of 92 (estimated)~1074 Elo
  1. #1Claude Opus 5.51866
  2. #2Muse Wonder Full1857
  3. #37Inkling1076
  4. ~#38Nanbeige4.1-3BEstimated1074
  5. #38Grok Build 0.1 06161052
No published score — estimated from related benchmarks
~#47 of 85 (estimated)~10.7%
  1. #1Claude Fable 5.276.4%
  2. #2Muse Wonder Full71.3%
  3. #46K2 Horizon 3.7B14.7%
  4. ~#47Nanbeige4.1-3BEstimated10.7%
  5. #47Qwen3.5 397B A17B6.9%
No published score — estimated from related benchmarks
~#57 of 91 (estimated)~12.6%
  1. #1Muse Spark 1.350.5%
  2. #2GLM-5.350.3%
  3. #56gpt-oss-120b12.8%
  4. ~#57Nanbeige4.1-3BEstimated12.6%
  5. #57Gemma 4 26B A4B12.0%
No published score — estimated from related benchmarks